Visit us at the poster session Tues, 10:45 am at #NeurIPS2023!
Explainability research has mainly focused on extracting explanations that improve user understanding. We take it a step further and ask: can explanations also benefit the AI agent and improve agent learning rate?
Overall, our work examines this interesting idea on the dual benefit of concept-based explanations to both agents as well as users!
Check out our paper for more details!
Paper: https://t.co/QXUWb6hK0a
Work w/ Sonia Chernova (@ICatGT) and @_beenkim
Visit us at the poster session Tues, 10:45 am at #NeurIPS2023!
Explainability research has mainly focused on extracting explanations that improve user understanding. We take it a step further and ask: can explanations also benefit the AI agent and improve agent learning rate?
Our user study results show that concept-based explanations do benefit end users and significantly improve user task performance compared to existing baselines for both Connect 4 and Lunar Lander.
Should we be worried about AI? @ICatGT's Sonia Chernova answers in our first mini-sode of the new series #TechTakes. Watch the full episode at https://t.co/Gg4pU2gH6t
Tell me why. XAI, I want it that way! Amazing music video on explainability in machine learning by @FinaleDoshi's lab. My cameo along with @QVeraLiao@zywind Ronny and Amit is at 3:04. https://t.co/gWaW5Ex7bW
None of the work I'll be presenting would have been possible without brilliant people who I was lucky enough to work with:
@devleenadas_, Sonia Chernova, Amy Widdicombe, Simon Julier, Mike Mozer, @PangWeiKoh, @thao_nguyen26, Yew Siang Tang, @MussmannSteve, @2plus2make5, 4/n